Abstract
This dissertation discusses motivations and methods for translation of knowledge bases (KBs) between different expert system building tools (ESBTs). A Knowledge Canonical Form (KCF) is presented as a generic knowledge representation mechanism. A set of KB translations between different ESBTs through the KCF is shown to reduce overhead. A set of fuzzy mapping functions address equivalence between different ESBT uncertainty management systems (UMSs). This dissertation prescribes methods for translation between certainty factor (CF) based UMSs and the scales used to represent uncertainty. Conversion between other UMSs and the uncertainty scales are also discussed. The major benefits from this research are twofold. First, the research demonstrates reuse of KBs in different environments. Second, the methodology from this work can be extended and reused. Some of the problems encountered during automated translation are explained with suggested solutions. A survey and analysis of eight commercial ESBTs focuses on the UMSs incorporated in each tool. This work has direct application in industry. Feasibility is demonstrated through a set of prototype translators which are currently being designed in C++. Conclusions and recommendations for extending the KCF to other domains are made.
Jilg, Jeffrey Michael (1992). A fuzzy uncertainty management translator. Texas A&M University. Texas A&M University. Libraries. Available electronically from
https : / /hdl .handle .net /1969 .1 /DISSERTATIONS -1397366.